Researchers Build AI Guard Against Robocalls

A virtual assistant may be the key to preventing the endless robocalls annoying people with attempted scams every day. A prototype demonstrated in a paper from researchers at the Georgia Institute of Technology successfully checked incoming calls and blocked those from bots and could be used as an app on smartphones to protect people from future spam calls.


People in the U.S. received 18 robocalls a month on average in 2019, and the number of calls rose 35%, suggesting it’s even higher now.. While blocklists and other regulation may stem some of the calls, there are always plenty more that will get through. Smartphones are getting better at identifying robocalls, and will warn a user of “Spam Risk” calls sometimes, but the mild annoyance of rejecting those calls can build up over time. The researchers suggested a way of stopping the robocalls before a human has to deal with them at all.

“Similar to a human assistant, the virtual assistant can pick up an incoming call and screen it without user interruption to determine if the call is unwanted,” the paper’s authors explained. “Via a user study, we show that our virtual assistant is able to preserve the user experience of a typical phone call. At the same time, we show that our system can detect mass robocalls without negatively impacting legitimate callers.”

The researchers designed an Android app named RobocallGuard to check the number of an incoming call against lists of approved and blacklisted numbers, so anyone in a user’s contact list gets through and any previously blocked number never appears. If the number is on neither list, which is often how robocalls keep redialing, RobocallGuard answers for the user and asks who they are calling for. If the name matches who the number belongs to, the call is sent through with a transcript of the conversation. If the caller fails, the number is blocked and the user is notified about what happened.

App Evolution

This isn’t the first attempt at designing an AI guard against robocalls. Google created a call screening service for its Pixel phones back in 2018, later adding transcription and other improvements to the feature. When the call arrives, the user can screen it and have the AI answer the phone, asking the caller questions from a limited setlist and sending the response to the phone owner to decide if they want the call to come through or not. What the GIT researchers have developed wouldn’t need any input from the phone’s owner at all.

RobocallGuard worked very effectively, blocking every one of around 8,000 robocalls in a test, and successfully identifying and letting through 97.8% of human callers. But that doesn’t mean it’s ready for general use. The authors point out that every attempt to limit robocalls is met in turn by a more sophisticated system from the groups running the robocalls. The question of who they are trying to call works well now, but a way of handling more devious robocallers will be crucial to make the solution stick in the long term. It’s an evolutionary tech battle that email platforms deal with regarding spam, for instance. Perhaps at some point, RoboCallGuard will block as effectively as most spam filters, but until then we’ll all be declining unknown numbers by hand.


Google Assistant’s AI Call Screening Feature to get Transcription Support in 2018

Raise to Talk to Google Assistant Feature Planned for Google Pixel 4: Report